static uint32_t readutf(char c, char **buf, size_t *buflen, size_t *sz) {
size_t i, len;
uint32_t val;
if ((c & 0x80) == 0)
len = 1;
else if ((c & 0xe0) == 0xc0)
len = 2;
else if ((c & 0xf0) == 0xe0)
len = 3;
else if ((c & 0xf8) == 0xf0)
len = 4;
else
lfatal(curloc, "Invalid utf8 encoded character constant");
val = c & ((1 << (8 - len)) - 1);
append(buf, buflen, sz, c);
for (i = 1; i < len; i++) {
c = next();
if ((c & 0xc0) != 0x80)
lfatal(curloc, "Invalid utf8 codepoint in character literal");
val = (val << 6) | (c & 0x3f);
append(buf, buflen, sz, c);
}
return val;
}
static Tok *charlit(void)
{
Tok *t;
int c;
uint32_t val;
size_t len, sz;
char *buf;
assert(next() == '\'');
buf = NULL;
len = 0;
sz = 0;
val = 0;
c = next();
if (c == End)
lfatal(curloc, "Unexpected EOF within char lit");
else if (c == '\n')
lfatal(curloc, "Newlines not allowed in char lit");
else if (c == '\\')
val = decode(&buf, &len, &sz);
else
val = readutf(c, &buf, &len, &sz);
append(&buf, &len, &sz, '\0');
if (next() != '\'')
lfatal(curloc, "Character constant with multiple characters");
t = mktok(Tchrlit);
t->chrval = val;
t->id = buf;
return t;
}
/*
* decodes an escape code. These are
* shared between strings and characters.
* Unknown escape codes are ignored.
*/
static int decode(char **buf, size_t *len, size_t *sz)
{
char c, c1, c2;
int32_t v;
c = next();
/* we've already seen the '\' */
switch (c) {
case 'u':
v = unichar();
appendc(buf, len, sz, v);
return v;
case 'x': /* arbitrary hex */
c1 = next();
if (!isxdigit(c1))
lfatal(curloc, "expected hex digit, got %c", c1);
c2 = next();
if (!isxdigit(c2))
lfatal(curloc, "expected hex digit, got %c", c1);
v = 16*hexval(c1) + hexval(c2);
break;
case 'n': v = '\n'; break;
case 'r': v = '\r'; break;
case 't': v = '\t'; break;
case 'b': v = '\b'; break;
case '"': v = '\"'; break;
case '\'': v = '\''; break;
case 'v': v = '\v'; break;
case '\\': v = '\\'; break;
case '0': v = '\0'; break;
default: lfatal(curloc, "unknown escape code \\%c", c);
}
append(buf, len, sz, v);
return v;
}
static Tok *strlit(void)
{
Tok *t;
int c;
size_t len, sz;
char *buf;
assert(next() == '"');
buf = NULL;
len = 0;
sz = 0;
while (1) {
c = next();
/* we don't unescape here, but on output */
if (c == '"')
break;
else if (c == End)
lfatal(curloc, "Unexpected EOF within string");
else if (c == '\n')
lfatal(curloc, "Newlines not allowed in strings");
else if (c == '\\')
decode(&buf, &len, &sz);
else
append(&buf, &len, &sz, c);
};
t = mktok(Tstrlit);
t->strval.len = len;
/* null terminator should not count towards length */
append(&buf, &len, &sz, '\0');
t->strval.buf = buf;
t->id = buf;
return t;
}
// ######################################################################
void env_stdio_write_gray(const struct env_image* iimage,
const char* outstem, const char* name, int c)
{
if (!env_img_initialized(iimage))
return;
// if outstem is the empty string, then the user wanted us to
// suppress output:
if (outstem[0] == '\0')
return;
char fname[256];
snprintf(fname, sizeof(fname),
"%s-%s%06d.pnm", outstem, name, c);
FILE* const f = fopen(fname, "wb");
if (f == 0)
lfatal("%s: couldn't open PNM file for writing (errno=%d, %s)",
fname, errno, strerror(errno));
//REV: so they are P5 (binary, greyscale), dim * dim images, with max value 255.
if (fprintf(f, "P5\n%d %d\n255\n",
(int) iimage->dims.w, (int) iimage->dims.h) < 0)
lfatal("%s: fprintf() failed (errno=%d, %s)",
fname, errno, strerror(errno));
const intg32* const src = env_img_pixels(iimage);
const env_size_t sz = env_img_size(iimage);
byte* bimage =
(byte*) env_allocate(sz * sizeof(byte));
for (env_size_t i = 0; i < sz; ++i)
{
// the caller is supposed to have already ensured that
// the intg32 image has been downsampled to a [0,255]
// range, so let's verify that:
ENV_ASSERT(src[i] >= 0 && src[i] <= 255);
bimage[i] = (byte) src[i];
}
//REV: they aren't using any newlines!
if (fwrite(bimage, 1, sz, f) != sz)
{
env_deallocate(bimage);
lfatal("%s: fwrite() failed (errno=%d, %s)",
fname, errno, strerror(errno));
}
env_deallocate(bimage);
if (fclose(f) != 0)
lfatal("%s: fclose() failed (errno=%d, %s)",
fname, errno, strerror(errno));
}
static void eatcomment(void)
{
int depth;
int startln;
int c;
depth = 0;
startln = curloc.line;
while (1) {
c = next();
switch (c) {
/* enter level of nesting */
case '/':
if (match('*'))
depth++;
break;
/* leave level of nesting */
case '*':
if (match('/'))
depth--;
break;
/* have to keep line numbers synced */
case '\n':
curloc.line++;
break;
case End:
lfatal(curloc, "File ended within comment starting at line %d", startln);
break;
}
if (depth == 0)
break;
}
}
static Tok *toknext()
{
Tok *t;
int c;
eatspace();
c = peek();
if (c == End) {
t = mktok(0);
} else if (c == '\n') {
curloc.line++;
next();
t = mktok(Tendln);
} else if (isalpha(c) || c == '_' || c == '$') {
t = kwident();
} else if (c == '"') {
t = strlit();
} else if (c == '\'') {
t = charlit();
} else if (isdigit(c)) {
t = numlit();
} else if (c == '@') {
t = typaram();
} else {
t = oper();
}
if (!t || t->type == Terror)
lfatal(curloc, "Unable to parse token starting with %c", c);
return t;
}
/* \u{abc} */
static int32_t unichar(void)
{
uint32_t v;
int c;
/* we've already seen the \u */
if (next() != '{')
lfatal(curloc, "\\u escape sequence without initial '{'");
v = 0;
while (ishexval(peek())) {
c = next();
v = 16*v + hexval(c);
if (v > 0x10FFFF)
lfatal(curloc, "invalid codepoint for \\u escape sequence");
}
if (next() != '}')
lfatal(curloc, "\\u escape sequence without ending '}'");
return v;
}
/*
* Converts a character to its hex value.
*/
static int hexval(char c)
{
if (c >= 'a' && c <= 'f')
return c - 'a' + 10;
else if (c >= 'A' && c <= 'F')
return c - 'A' + 10;
else if (c >= '0' && c <= '9')
return c - '0';
lfatal(curloc, "passed non-hex value '%c' to where hex was expected", c);
return -1;
}
static void fixtypemappings(Stab *st)
{
size_t i;
Type *t, *old;
/*
* merge duplicate definitions.
* This allows us to compare named types by id, instead
* of doing a deep walk through the type. This ability is
* depended on when we do type inference.
*/
for (i = 0; i < ntypefixdest; i++) {
t = htget(tidmap, itop(typefixid[i]));
if (!t)
die("Unable to find type for id %zd\n", typefixid[i]);
*typefixdest[i] = t;
}
for (i = 0; i < ntypefixdest; i++) {
old = *typefixdest[i];
if (old->type == Tyname || old->type == Tygeneric) {
t = htget(tydedup, old);
if (!t) {
t = old;
htput(tydedup, old, old);
}
*typefixdest[i] = t;
}
}
/* check for duplicate type definitions */
for (i = 0; i < ntypefixdest; i++) {
t = htget(tidmap, itop(typefixid[i]));
if ((t->type != Tyname && t->type != Tygeneric) || t->issynth)
continue;
old = htget(tydedup, t);
if (old && !tyeq(t, old) && !isspecialization(t, old))
lfatal(t->loc, "Duplicate definition of type %s on %s:%d", tystr(old), file->file.files[old->loc.file], old->loc.line);
}
for (i = 0; i < ntypefixdest; i++)
lfree(&typefixdest, &ntypefixdest);
lfree(&typefixid, &ntypefixid);
}
/*
* Appends a unicode codepoint 'c' to a growable buffer 'buf',
* resizing if needed.
*/
static void appendc(char **buf, size_t *len, size_t *sz, uint32_t c)
{
size_t i, charlen;
char charbuf[5] = {0};
if (c < 0x80)
charlen = 1;
else if (c < 0x800)
charlen = 2;
else if (c < 0x10000)
charlen = 3;
else if (c < 0x200000)
charlen = 4;
else
lfatal(curloc, "invalid utf character '\\u{%x}'", c);
encode(charbuf, charlen, c);
for (i = 0; i < charlen; i++)
append(buf, len, sz, charbuf[i]);
}
static Tok *number(int base)
{
Tok *t;
int start;
int c;
int isfloat;
int unsignedval;
/* because we allow '_' in numbers, and strtod/stroull don't, we
* need a buffer that holds the number without '_'.
*/
char buf[2048];
size_t nbuf;
t = NULL;
isfloat = 0;
start = fidx;
nbuf = 0;
for (c = peek(); isxdigit(c) || c == '.' || c == '_'; c = peek()) {
next();
if (c == '_')
continue;
if (c == '.')
isfloat = 1;
else if (hexval(c) < 0 || hexval(c) > base)
lfatal(curloc, "Integer digit '%c' outside of base %d", c, base);
if (nbuf >= sizeof buf - 1) {
buf[nbuf-1] = '\0';
lfatal(curloc, "number %s... too long to represent", buf);
}
buf[nbuf++] = c;
}
buf[nbuf] = '\0';
/* we only support base 10 floats */
if (isfloat && base == 10) {
t = mktok(Tfloatlit);
t->id = strdupn(&fbuf[start], fidx - start);
t->fltval = strtod(buf, NULL);
} else {
t = mktok(Tintlit);
t->id = strdupn(&fbuf[start], fidx - start);
t->intval = strtoull(buf, NULL, base);
/* check suffixes:
* u -> unsigned
* l -> 64 bit
* i -> 32 bit
* w -> 16 bit
* b -> 8 bit
*/
unsignedval = 0;
nextsuffix:
switch (peek()) {
case 'u':
if (unsignedval == 1)
lfatal(curloc, "Duplicate 'u' integer specifier");
next();
unsignedval = 1;
goto nextsuffix;
case 'l':
next();
if (unsignedval)
t->inttype = Tyuint64;
else
t->inttype = Tyint64;
break;
case 'i':
next();
if (unsignedval)
t->inttype = Tyuint32;
else
t->inttype = Tyint32;
break;
case 's':
next();
if (unsignedval)
t->inttype = Tyuint16;
else
t->inttype = Tyint16;
break;
case 'b':
next();
if (unsignedval)
t->inttype = Tyuint8;
else
t->inttype = Tyint8;
break;
default:
if (unsignedval)
lfatal(curloc, "Unrecognized character int type specifier after 'u'");
break;
}
}
return t;
}
static Tok *oper(void)
{
int tt;
char c;
c = next();
switch (c) {
case '{': tt = Tobrace; break;
case '}': tt = Tcbrace; break;
case '(': tt = Toparen; break;
case ')': tt = Tcparen; break;
case '[': tt = Tosqbrac; break;
case ']': tt = Tcsqbrac; break;
case ',': tt = Tcomma; break;
case '`': tt = Ttick; break;
case '#': tt = Tderef; break;
case ':':
if (match(':'))
tt = Twith;
else
tt = Tcolon;
break;
case '~': tt = Tbnot; break;
case ';':
if (match(';'))
tt = Tendblk;
else
tt = Tendln;
break;
case '.':
if (match('.')) {
if (match('.')) {
tt = Tellipsis;
} else {
unget();
tt = Tdot;
}
} else {
tt = Tdot;
}
break;
case '+':
if (match('='))
tt = Taddeq;
else if (match('+'))
tt = Tinc;
else
tt = Tplus;
break;
case '-':
if (match('='))
tt = Tsubeq;
else if (match('-'))
tt = Tdec;
else if (match('>'))
tt = Tret;
else
tt = Tminus;
break;
case '*':
if (match('='))
tt = Tmuleq;
else
tt = Tmul;
break;
case '/':
if (match('='))
tt = Tdiveq;
else
tt = Tdiv;
break;
case '%':
if (match('='))
tt = Tmodeq;
else
tt = Tmod;
break;
case '=':
if (match('='))
tt = Teq;
else
tt = Tasn;
break;
case '|':
if (match('='))
tt = Tboreq;
else if (match('|'))
tt = Tlor;
else
tt = Tbor;
break;
case '&':
if (match('='))
tt = Tbandeq;
else if (match('&'))
tt = Tland;
else
tt = Tband;
break;
case '^':
if (match('='))
tt = Tbxoreq;
else
tt = Tbxor;
break;
case '<':
if (match('=')) {
tt = Tle;
} else if (match('<')) {
if (match('='))
tt = Tbsleq;
else
tt = Tbsl;
} else {
tt = Tlt;
}
break;
case '>':
if (match('=')) {
tt = Tge;
} else if (match('>')) {
if (match('='))
tt = Tbsreq;
else
tt = Tbsr;
} else {
tt = Tgt;
}
break;
case '!':
if (match('='))
tt = Tne;
else
tt = Tlnot;
break;
default:
tt = Terror;
lfatal(curloc, "Junk character %c", c);
break;
}
return mktok(tt);
}
// ######################################################################
struct env_rgb_pixel* env_stdio_parse_rgb(const char* fname,
struct env_dims* outdims)
{
FILE* f = fopen(fname, "rb");
if (f == 0)
lfatal("Couldn't open file '%s' for reading.", fname);
int c = getc(f);
if (c != 'P')
lfatal("Missing magic number in pnm file '%s'"
"(got '%c' [%d], expected '%c' [%d]).",
fname, c, c, 'P', 'P');
int mode = -1;
int ret = fscanf(f, "%d", &mode);
if (ret > 0 && mode != 6)
lfatal("Wrong pnm mode (got 'P%d', expected 'P6')", //REV: so they expect P6, i.e. binary rgb!
mode);
while (1)
{
const int c = getc(f); //REV: they're getting one char at a time...removing all spaces...
if (!isspace(c))
{ ungetc(c, f); break; }
}
// copy and concatenate optional comment line(s) starting with '#'
// into comments string
//REV: they're removing comments and such...we won't need to worry about this...
while (1)
{
const int c = getc(f);
if (c != '#')
{ ungetc(c, f); break; }
else
{
while (getc(f) != '\n')
{ /* empty loop */ }
}
}
int w = -1;
int h = -1;
int maxGrey = -1;
ret = fscanf(f, "%d %d %d", &w, &h, &maxGrey); //REV: they get the dimensions... and the max grey value (wth?)
ENV_ASSERT(ret > 0);
ENV_ASSERT(w > 0);
ENV_ASSERT(h > 0);
ENV_ASSERT(maxGrey > 0);
// read one more character of whitespace from the stream after maxGrey
c = getc(f);
if ( !isspace(c) )
lfatal("Missing whitespace after maxGrey in pbm file '%s'.", fname);
struct env_rgb_pixel* result = (struct env_rgb_pixel*) //REV: note RGB pixel is byte[3].
env_allocate(w * h * sizeof(struct env_rgb_pixel));
if (fread((char*) result, 3, w*h, f) != ((env_size_t)(w*h)))
lfatal("%s: fread() failed", fname);
outdims->w = w;
outdims->h = h;
return result;
}
